GCN_classification/usage_AM.txt at master · AndresPMD/GCN_classification · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
------- CONTEXT -------
TRAIN:
python train.py context --batch_size 64 --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat --split 1

TEST EVAL:
python train.py context --batch_size 64 --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat --split 1 --save_weights False --load ../best/context_fullGCN_bboxes_fasttext_google_ocr_concat_mean_split1/checkpoint_context.weights --test True

EXTRACT FEATS:
python train.py context --batch_size 64 --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat --split 1 --save_weights False --load ../best/context_fullGCN_bboxes_fasttext_google_ocr_concat_mean_split1/checkpoint_context.weights --test True --save_img_feats True

PERFORMANCE TEST:
python performance.py context --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat --load ../best/context_fullGCN_bboxes_fasttext_google_ocr_concat_mean_split1/checkpoint_context.weights

------- DRINK BOTTLE -------
TRAIN:
python train.py bottles --batch_size 64 --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat --split 2

TEST EVAL:
python train.py bottles --batch_size 64 --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat --split 2 --save_weights False --load ../best/bottles_fullGCN_bboxes_fasttext_google_ocr_concat_mean_split2/checkpoint_bottles.weights --test True

EXTRACT FEATS:
python train.py bottles --batch_size 64 --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat --split 2 --save_weights False --load ../best/bottles_fullGCN_bboxes_fasttext_google_ocr_concat_mean_split2/checkpoint_bottles.weights --test True --save_img_feats True

PERFORMANCE TEST:
python performance.py bottles --model fullGCN_bboxes --ocr google_ocr --embedding fasttext --max_textual 15 --max_visual 36 --projection_layer mean --fusion concat